Re: [RFC] Autoload-breakpoints new version [3/9] notification async

[Hui Zhu <teawater at gmail dot com>]

On Wed, Aug 8, 2012 at 9:24 PM, Yao Qi <yao@codesourcery.com> wrote:
> On Tuesday, August 07, 2012 03:09:46 PM Hui Zhu wrote:
>> This patch is the extend for the notification function.
>> Current notification function cannot handle the packet when GDB doesn't want
>> to send or receive packet from the remote target. This patch do the extend
>> make GDB can handle the notification even if it doesn't send or receive
>> packet.
>
> I am trying to understand what current GDB can do, and what patched GDB can
> do.  If I understand it correctly,
>
>   - current GDB only handle notification when it sends or receives packets
>   - patched GDB can handle notification at any time when notification comes
> in.
>
> Looks this patch changes the way of GDB handling notifications (from "only-on-
> send-receive" handling to "anytime" handling), but has nothing to do with how
> notification is sent out from remote stub.  If so, why do you add
> "NotificationAsync" in remote_supported_packet?  Remote supported packets are
> used to tell GDB what feature remote stub has or doesn't have.  In other
> words, "NotificationAsync" is a feature of GDB, instead of remote stub, so we
> don't need "NotificationAsync".

I agree with you.


>
> I applied patch 1/9 and 3/9, and get following regressions in gdbserver with
> async mode on.
>
>   FAIL: gdb.mi/mi-nonstop-exit.exp: finished exec continue (unknown output
> after running)
>   FAIL: gdb.mi/mi-nonstop-exit.exp: finished exec continue (2) (unknown output
> after running)
>   FAIL: gdb.mi/mi-nsmoribund.exp: resume all, waiting for program exit
> (timeout)

My part is OK, Could you tell me how do you do the test?

>
> --
> Yao (éå)

Thanks,
Hui